Extrinsic Cooperation Ceiling Theory

• Extrinsic Cooperation Ceiling is a concept defining externally imposed boundaries that limit the attainable level and stability of cooperative behavior in various systems.
• It encompasses multiple formulations—from strict upper bounds like pC ≤ ½ in population dynamics to threshold-based participation filters in success-driven group formation—highlighting diverse mechanisms of external constraint.
• Analytical models, simulations, noise-driven dynamics, and strategic game-theoretic approaches illustrate how these ceilings shape cooperation in fields ranging from evolutionary biology to network communications.

Searching arXiv for the cited papers and closely related work on cooperation ceilings, extrinsic constraints, and success-driven group formation. First, I’ll retrieve the principal paper on success-driven group formation and then complementary work on explicit “cooperation ceiling” formulations and related extrinsic mechanisms. Extrinsic cooperation ceiling denotes an externally imposed boundary on the attainable level, stability, or enactment of cooperation. In the studies summarized here, it is not a single standardized observable. It appears as a hard upper bound on the steady-state mean proportion of cooperators, as a threshold window created by an external participation filter, as a bound on cooperator fixation probability in fluctuating environments, and as a benchmark-relative cap on feasible cooperative surplus in one-shot games (Foster et al., 30 Jun 2026, Szolnoki et al., 2016, Assaf et al., 2013, Rong et al., 2014).

1. Definitions and major formulations

The term is used most sharply when an external mechanism, rather than intrinsic prosociality, limits cooperative outcomes. In heterogeneous evolutionary dynamics, the ceiling can be an explicit theorem: for a broad class of purely extrinsic update rules, the steady-state mean proportion of cooperators satisfies $p_C \le \frac12$ (Foster et al., 30 Jun 2026). In spatial public-goods games with success-driven group formation, the external threshold $H$ is not a simple hard cap on cooperation; it is an extrinsic participation filter that produces an optimal intermediate regime and a frozen state when the threshold is too strict (Szolnoki et al., 2016). In finite-population prisoner's dilemma under extrinsic noise, the relevant bounded quantity is not a stationary cooperation fraction but the cooperator fixation probability $\phi_{\mathrm C}(x_0)$, which is exponentially small in a static environment and can become only algebraically suppressed under strong short-correlated environmental fluctuations (Assaf et al., 2013). In cooperative-equilibrium theory, the closest formal analogue is the maximal common surplus above the best-response benchmark, $\alpha^*=\max_s \min_i(U_i(s)-BU_i^G)$, attained by M-PCE (Rong et al., 2014).

Formulation	Bounded quantity	Characteristic boundary
Purely extrinsic population dynamics	Steady-state mean proportion of cooperators $p_C$	$p_C \le \frac12$ (Foster et al., 30 Jun 2026)
Success-driven group formation	Cooperation under thresholded group initiation	Optimal intermediate $H$; too high $H$ yields frozen state $F$ (Szolnoki et al., 2016)
Finite PD under extrinsic noise	Cooperator fixation probability $\phi_{\mathrm C}(x_0)$	Static $H$0; strong EN gives algebraic suppression (Assaf et al., 2013)
M-PCE / cooperative equilibrium	Uniform cooperative surplus above benchmark	$H$1 (Rong et al., 2014)

A plausible synthesis is that the phrase names a family of externally generated boundary conditions rather than a universal law. Some formulations cap realized cooperation directly, some bound only its attainability, and some define the boundary relative to strategic benchmarks rather than raw frequency.

2. Hard ceilings from purely extrinsic population dynamics

The most explicit formal ceiling is given by the distinction between purely extrinsic and purely intrinsic population dynamics in heterogeneous populations (Foster et al., 30 Jun 2026). A transition matrix is purely extrinsic when neighboring transitions depend only on payoffs in the current state, are nondecreasing in the payoffs of current role models already using the candidate action, are nonincreasing in the payoffs of non-role-models, and do not directly depend on the labels $H$2 apart from focal payoff terms, mutation, and the definition of the role-model set. Moran and Fermi imitation are the canonical extrinsic rules. Purely intrinsic dynamics depend only on the focal player's own payoff criterion; aspiration dynamics and introspection are the canonical intrinsic rules.

The theorem labeled “Extrinsic Ceiling” states that, for a neutrally monotone purely extrinsic process on an $H$3-player game with two actions $H$4, if in every state $H$5,

$H$6

then under action-invariant mutation the steady-state mean proportion of cooperators satisfies

$H$7

The bounded quantity is

$H$8

This is an analytical theorem, not only a simulation regularity (Foster et al., 30 Jun 2026).

The paper demonstrates the mechanism in a heterogeneous public-goods game where player $H$9's payoff is

$\phi_{\mathrm C}(x_0)$0

Holding everyone else fixed,

$\phi_{\mathrm C}(x_0)$1

Hence cooperation is individually better when $\phi_{\mathrm C}(x_0)$2. Yet within a given mixed state every defector still avoids its own contribution cost, so outward-looking dynamics continue to favor defection. This is why the ceiling is specific to extrinsic updating rather than to the game itself.

The numerical sweep reinforces the theorem. Across $\phi_{\mathrm C}(x_0)$3 parameter sets, Moran and Fermi exceeded $\phi_{\mathrm C}(x_0)$4 in $\phi_{\mathrm C}(x_0)$5 of cases, while introspection and aspiration exceeded $\phi_{\mathrm C}(x_0)$6 in $\phi_{\mathrm C}(x_0)$7 overall and $\phi_{\mathrm C}(x_0)$8 for $\phi_{\mathrm C}(x_0)$9. Introspection crosses the neutral baseline exactly at $\alpha^*=\max_s \min_i(U_i(s)-BU_i^G)$0. The paper therefore calls the excess above $\alpha^*=\max_s \min_i(U_i(s)-BU_i^G)$1 an “intrinsic escape”: the ceiling is generated by the outward-looking rule, not by the heterogeneous public-goods environment alone (Foster et al., 30 Jun 2026).

3. Participation filters and threshold windows

A different formulation appears in “success-driven group formation,” a spatial public-goods game on a square lattice with a von Neumann neighborhood, where each player belongs to $\alpha^*=\max_s \min_i(U_i(s)-BU_i^G)$2 possible groups and a player may organize a group only if it was sufficiently successful in the previous round (Szolnoki et al., 2016). The model introduces four states, $\alpha^*=\max_s \min_i(U_i(s)-BU_i^G)$3, $\alpha^*=\max_s \min_i(U_i(s)-BU_i^G)$4, $\alpha^*=\max_s \min_i(U_i(s)-BU_i^G)$5, and $\alpha^*=\max_s \min_i(U_i(s)-BU_i^G)$6, where high-merit players may organize a game and low-merit players may not. The threshold rule is strategy-neutral at the rule level but strategy-asymmetric in its consequences.

Merit is assigned stochastically by

$\alpha^*=\max_s \min_i(U_i(s)-BU_i^G)$7

with $\alpha^*=\max_s \min_i(U_i(s)-BU_i^G)$8, and strategy imitation uses

$\alpha^*=\max_s \min_i(U_i(s)-BU_i^G)$9

again with $p_C$0. Only groups centered on high-merit players are active. Low-merit players may still participate in neighbors’ active groups, but they cannot initiate one themselves. The standard public-goods interaction remains unchanged: cooperators contribute $p_C$1, defectors contribute nothing, total contribution is multiplied by synergy factor $p_C$2, and proceeds are divided equally among the $p_C$3 members.

This produces an extrinsic participation filter rather than a universal hard cap. If $p_C$4 is too low, the filter is weak and the model behaves essentially like the standard spatial public-goods game. If $p_C$5 is in an intermediate range, cooperation rises sharply because cooperators embedded in cooperative neighborhoods can repeatedly satisfy the threshold more sustainably than defectors. If $p_C$6 is too high, high-merit players vanish, no groups are formed, and the system enters the frozen state $p_C$7, where all high-merit players die out early and no proper public-goods interactions remain (Szolnoki et al., 2016).

The quantitative phase structure is specific. In the traditional model on the square lattice, coexistence begins only above

$p_C$8

For $p_C$9, which would yield full defection in the traditional model, success-driven group formation changes the outcome if the threshold exceeds

$p_C \le \frac12$0

Above this value cooperators and defectors coexist; cooperation keeps increasing with $p_C \le \frac12$1, and for

$p_C \le \frac12$2

the system reaches a full-cooperator state, before still larger $p_C \le \frac12$3 pushes the system into the frozen state. For $p_C \le \frac12$4, cooperation is improved above

$p_C \le \frac12$5

and a full-cooperator phase is reached for

$p_C \le \frac12$6

and above, before freezing again appears at still larger $p_C \le \frac12$7.

The model also shows that the synergy factor can itself enter a bounded window. At $p_C \le \frac12$8, cooperation can already be supported at

$p_C \le \frac12$9

well below the classical coexistence threshold $H$0. Yet increasing $H$1 from there may initially reduce cooperation because a larger synergy factor allows defectors too to surpass the threshold and become organizers. This makes the notion of a ceiling more precise: the external threshold is useful only inside a threshold window, and under strict thresholds there can also be an optimal synergy window. The mechanism is robust on a random graph with uniform degree distribution and mean degree $H$2, where for $H$3 cooperation begins to improve above $H$4 and freezes again for $H$5. By contrast, in the well-mixed version the mechanism does not actively promote cooperation; low $H$6 gives full defection, and high $H$7 eliminates both high-merit types and leaves only low-merit players with no actual interactions (Szolnoki et al., 2016).

4. Environmental, topological, and allocation-based modifiers

Extrinsic environmental fluctuations can relax a severe cooperation ceiling even when deterministic selection favors defection. In a finite well-mixed prisoner's dilemma with Moran-type birth-death updating, the static-environment fixation probability is

$H$8

This is the paper’s static baseline. Environmental fluctuations are introduced by

$H$9

where $H$0 is Ornstein-Uhlenbeck noise with

$H$1

Under strong short-correlated extrinsic noise, the suppression of cooperation changes from exponential to algebraic: $H$2 Under strong adiabatic noise, the environment becomes the limiting timescale and

$H$3

This is a ceiling on attainability rather than on the stationary cooperation fraction, but it is still extrinsic because the limiting barrier is set by environmental variability (Assaf et al., 2013).

Interaction topology can itself generate external bottlenecks. In a network public-goods game on a $H$4-regular graph, the linear case requires

$H$5

for cooperation to emerge or stabilize. Under discounting, cooperation becomes systematically harder as $H$6 increases. Under weak synergy, the critical threshold becomes non-monotonic in $H$7: a moderate number of neighbors is worst, and “cooperation with both synergistic and local interactions can be worse than each alone.” The paper’s phrase “synergistic interactions work with strangers but not with neighbors” makes the ceiling interpretation explicit: externally imposed local interaction structure can turn two individually cooperation-promoting mechanisms into a suppressive regime (Li et al., 2014).

A third mechanism changes the ceiling by reallocating rather than increasing cooperative effort. In a spatial public-goods game on a square lattice with $H$8 overlapping groups, a selective cooperator still contributes $H$9 to her own group but allocates the remaining external budget $F$0 exclusively to the game organized by the neighboring player with the highest payoff. Total contribution remains conserved. Under the traditional model, conditions such as

$F$1

lead to full defection. With a sufficient fraction of selective players, the same setting can reach full cooperation. The effect is robust across deterministic, uniform, bimodal, truncated exponential, and truncated power-law assignments of selective propensity. The main spatial condition is that selective players are not isolated: once their supporting influence percolates, cooperative regions overlap and the ordinary ceiling is sharply lifted (Lee et al., 2021).

5. Strategic and thermodynamic interpretations

In strategic-form game theory, the nearest formal analogue of an extrinsic cooperation ceiling is not a frequency bound but a benchmark-relative cap on cooperative surplus. In two-player games, perfect cooperative equilibrium uses

$F$2

and an $F$3-PCE requires

$F$4

The maximal feasible $F$5 is

$F$6

and an M-PCE attains this value. This is a ceiling because no strategy profile can make every player exceed the best-response benchmark by more than $F$7. In the Prisoner’s Dilemma, $F$8 is a $F$9-PCE and the unique M-PCE; in the Nash bargaining game the unique M-PCE is $\phi_{\mathrm C}(x_0)$0 and it is a $\phi_{\mathrm C}(x_0)$1-PCE. The ceiling is therefore on uniform cooperative surplus above an externally determined benchmark, not on welfare simpliciter (Rong et al., 2014).

Capraro’s cooperative equilibrium uses an explicitly payoff-sensitive coalition forecast. For coalition structure $\phi_{\mathrm C}(x_0)$2, the prior probability that player $\phi_{\mathrm C}(x_0)$3 deviates is

$\phi_{\mathrm C}(x_0)$4

and the coalition value is

$\phi_{\mathrm C}(x_0)$5

The induced game $\phi_{\mathrm C}(x_0)$6 then keeps only mixed-strategy profiles $\phi_{\mathrm C}(x_0)$7 such that

$\phi_{\mathrm C}(x_0)$8

This makes cooperation explicitly dependent on external temptation, strategic risk, and coalition-breakdown probabilities. In the parametrized Prisoner’s Dilemma $\phi_{\mathrm C}(x_0)$9, the unique cooperative equilibrium is $H$00 if $H$01 and a mixed profile with cooperation probability $H$02 if $H$03. In Traveler’s Dilemma, the cooperative coalition value decreases with the bonus/penalty $H$04. In the public-goods game, $H$05 is strictly increasing in the marginal return $H$06, negative at $H$07, and positive at $H$08. These are payoff-driven ceilings in a direct sense (Capraro, 2013).

A thermodynamic-limit version appears in the Ising mapping of symmetric $H$09 games. There the order parameter is

$H$10

with $H$11 and $H$12 determined by payoff differences. In the Prisoner’s Dilemma,

$H$13

under the standard ordering $H$14, so magnetization is typically negative if cooperation is mapped to $H$15; the paper states that no phase transition to cooperative majority occurs in the valid Prisoner’s Dilemma regime. In the Game of Chicken, the majority switches at

$H$16

Here the ceiling is a thermodynamic-limit restriction: cooperation can persist, but its dominance is controlled by external field-like payoff asymmetries and by temperature/noise, not by two-player Nash analysis alone (Benjamin et al., 2018).

6. Misconceptions, counterexamples, and broader usage

A common misconception is that any extrinsic pressure must cap cooperation from above. A direct counterexample is the spatial birth-death model in which a cooperator lowers the death rate of its direct neighbors while paying an increase in its own death rate proportional to the number of its neighbors. There the extrinsic control parameter is the baseline mortality $H$17. When benefit is strictly larger than cost, increasing $H$18 splits the ordinary contact-process extinction transition into

$H$19

and “full cooperation is established at the extinction transition as long as benefit is strictly larger than cost.” In that setting, extrinsic pressure acts as an external selector favoring cooperation rather than as a hard ceiling (Klemm et al., 2019).

Another misconception is that better general capability should eliminate cooperation ceilings. In a multi-agent LLM benchmark designed so that helping is strategically trivial and privately costless, the implemented perfect-play baseline achieved $H$20 tasks, yet baseline model performance ranged from $H$21 tasks for GPT-4.1-mini to $H$22 for Gemini-2.5-Pro, and capability did not predict cooperation. Explicit protocols and a tiny sender-side bonus of $H$23 per truthful send moved performance substantially, but typically not to perfect play. The paper therefore describes a movable but persistent ceiling created by cooperation, competence, and communication bottlenecks rather than by helper-side cost (Yadav et al., 9 Apr 2026).

The phrase also has cross-domain uses that are not about strategic cooperation among social agents. In large wireless networks, “Fundamental Limits of Cooperation” identifies a power-independent spectral-efficiency ceiling caused by out-of-cluster interference and finite coherence; even full transmitter cooperation cannot in general turn an interference-limited network into a noise-limited network, and spectral efficiency saturates at a finite $H$24 (Lozano et al., 2012). In electroaerodynamic propulsion, ceiling proximity can act as a passive cooperative element through modified inlet aerodynamics and plasma-mediated electrostatic attraction, producing up to $H$25 efficiency improvement in a small-scale thruster near a ceiling plane (Nelson et al., 2024). These engineering usages are terminologically related but conceptually distinct: the “ceiling” is literal or network-extrinsic rather than an upper bound on prosocial behavior.

Taken together, the literature does not support a single universal doctrine of an extrinsic cooperation ceiling. It supports a more differentiated view. Some extrinsic mechanisms impose a strict bound, such as $H$26 under purely extrinsic population dynamics. Some create an optimal window, as with success-driven group formation and threshold $H$27. Some relax a previously severe ceiling, as with extrinsic noise in finite populations or selective external investment in spatial public-goods games. Others show that external pressure can even select for full cooperation near extinction. The concept is therefore best understood as a family of externally generated boundary conditions on cooperation rather than a single fixed cap.